The Research Of Phishing Detection Technology Based On Nearest Neighbor And Similarity Measurement

As the development of the Internet and electronic commerce, more and more people play their economic activities through the network.The Internet and electronic commercial platform is not only making people convenient but also giving the phishing attackers to provide the platform of the illegal crime activity. The attacker design the phishing which is very similar to the legitimate web.Once the victims log in the phishing web and import their bank account and password and private information, the attackers will get the victims'private information. In recent years the phishing web is increasingly fierce and become the biggest problems which the Internet users have to face to. Thus it is becoming a top priority of all the Internat problems that developing a highly efficient phishing web detection technology for the workers of the network security.If the phishers try to lure users to log on the phishing web, they must make the phishing web pages look the same as the corresponding legal web pages. In view of this situation, this thesis firstly summarizes the fishing web detection method, and then puts forward the improvement based on the existing heuristic analysis of fishing web detection method: firstly, a new algorithm is proposed, which is a scale invariant feature transform algorithm based on the Heuristic analysis of the fishing web detection methods; Secondly, the thesis put forth the bayesian threshold value estimation model on the threshold estimation. the proposed method is tested through theoretical analysis and emulational experiments. The results show that the feature extraction of sift algorithm increases the accurate rate and the recall rate of the fishing web detection; especially the bayesian threshold value estimation model reduces the false negative rate. The two proposed methods not only increase the effectiveness and accuracy of the phishing web detection, but also improve the possibility of practical application.